Guaranteed Reach-Avoid for Black-Box Systems through Narrow Gaps via Neural Network Reachability

ICRA 2025

Long Kiu Chung¹, Wonsuhk Jung¹, Srivatsank Pullabhotla¹, Parth Shinde¹, Yadu Sunil¹, Saihari Kota¹, Luis Felipe Wolf Batista², Cédric Pradalier², Shreyas Kousik¹

¹Georgia Institute of Technology, ²Georgia Tech Europe

• Link
• GitHub
• YouTube

All of PARC‘s strengths, none of its weaknesses

---

Overview Video

---

Control Conservativeness with Network Size

We learn the robot’s trajectories as a neural network, allowing us to perform tighter, but still guaranteed safe maneuvers by simply increasing the network size.

---

Reach-Avoid Guarantees on Reinforcement Learning

NeuralPARC is agnostic to how the black-box robot generates its trajectories, and can therefore provide safety guarantees even on learning-based policies such as reinforcement learning (RL). We demonstrate this on an autonomous boat subjected to large disturbances.

---

Drift-Parking in Real Life

We validated NeuralPARC on a F1/10 model car performing drift parallel parking maneuvers.

---

Abstract

In the classical reach-avoid problem, autonomous mobile robots are tasked to reach a goal while avoiding obstacles. However, it is difficult to provide guarantees on the robot’s performance when the obstacles form a narrow gap and the robot is a black-box (i.e. the dynamics are not known analytically, but interacting with the system is cheap). To address this challenge, this paper presents NeuralPARC. The method extends the authors’ prior Piecewise Affine Reach-avoid Computation (PARC) method to systems modeled by rectified linear unit (ReLU) neural networks, which are trained to represent parameterized trajectory data demonstrated by the robot. NeuralPARC computes the reachable set of the network while accounting for modeling error, and returns a set of states and parameters with which the black-box system is guaranteed to reach the goal and avoid obstacles. Through numerical experiments, NeuralPARC is shown to outperform PARC in generating provably-safe extreme vehicle drift parking maneuvers, as well as enabling safety on an autonomous surface vehicle (ASV) subjected to large disturbances and controlled by a deep reinforcement learning (RL) policy.

---

BibTeX

@article{chung2025guaranteed,
    title   = {Guaranteed Reach-Avoid for Black-Box Systems through Narrow Gaps via Neural Network Reachability},
    author  = {Chung, Long Kiu and Jung, Wonsuhk and Pullabhotla, Srivatsank and Shinde, Parth and Sunil, Yadu and Kota, Saihari and Batista, Luis Felipe Wolf and Pradalier, C{\'e}dric and Kousik, Shreyas},
    booktitle={2025 IEEE International Conference on Robotics and Automation (ICRA)},
    year    = {2025},
    organization={IEEE}
}